CAJ | 학술논문

非凸在线支持向量机（LASVM-NC）具有抗噪能力强和训练速度快的优点，而词频相关频率积（tf.rf）则是一种自适应能力很强、分类性能非常好的文本特征。通过把非凸在线支持向量机和词频相关频率积相结合，提出了一种新的文本分类方法，即LASVM-NC+tf.rf。实验结果表明，这种方法在LASVM-NC与多种其他特征的结合中性能是最好的，且与SVM+tf.rf相比，不仅所产生的分类器具有泛化能力更强、模型表达更稀疏的优点，而且在处理含噪声的数据时具有更好的鲁棒性，在处理大规模数据时具有快得多的训练速度。
비철재선지지향량궤（LASVM-NC）구유항조능력강화훈련속도쾌적우점，이사빈상관빈솔적（tf.rf）칙시일충자괄응능력흔강、분류성능비상호적문본특정。통과파비철재선지지향량궤화사빈상관빈솔적상결합，제출료일충신적문본분류방법，즉LASVM-NC+tf.rf。실험결과표명，저충방법재LASVM-NC여다충기타특정적결합중성능시최호적，차여SVM+tf.rf상비，불부소산생적분류기구유범화능력경강、모형표체경희소적우점，이차재처리함조성적수거시구유경호적로봉성，재처리대규모수거시구유쾌득다적훈련속도。
Non-convex online support vector machine(LASVM-NC)has the advantages of strong anti-noise ability and fast training speed, while term frequency relevance frequency product(tf.rf)is a very good text feature for adaptive classi-fication performance. LASVM-NC+tf.rf is proposed as a new text classification method by combining non-convex support vector machines with term frequency relevance frequency product. It has been shown that the method can perform better than LASVM-NC plus many other features. Moreover, the method can produce faster trained and more robust classifiers with greater generalization and sparser expression than SVM+tf.rf in processing noisy and large-scale datasets.